Website Traffic Forecasting Using Deep Learning Techniques

Abstract

Today, predicting website traffic is a huge concern since it could influence the operation of critical websites. With more people visiting the website, it can crash or load very slowly. Such interruptions could resultin numerous disturbances. Users’ ratings of the site have subsequently dropped, and users have switched to other sites, which has an impact on the business. Web traffic is the volume of data that visitors send and receive on a website, and historically, it has made up the majority of internet traffic. The ability to forecast internet traffic flow is totally dependent on historical and real-time traffic data collected from many sources that monitor network flow. One of the toughest issues in this area is the prediction of future time series values. However, there are already a lot of systems and models for predicting internet traffic flow,the majority of them use basic traffic models and are still not completely satisfactory. This motivates us to revisit the deep learning-based model for predicting internet traffic flow given the abundance of available internet traffic data. The dataset used in this study contains information on the Hour Index and Sessions, and it was fed into CNN and LSTM time series forecasting models. The study compares the significance of the differences in the model to determine which makes better predictions for the traffic over the following 24 hours.